Epigenetic Regulation of the PTEN–AKT–RAC1 Axis by G9a Is Critical for Tumor Growth in Alveolar Rhabdomyosarcoma Akshay V

Published OnlineFirst March 4, 2019; DOI: 10.1158/0008-5472.CAN-18-2676

Cancer Molecular Cell Biology Research

Epigenetic Regulation of the PTEN–AKT–RAC1 Axis by G9a Is Critical for Tumor Growth in Alveolar Rhabdomyosarcoma Akshay V. Bhat1, Monica Palanichamy Kala1, Vinay Kumar Rao1, Luca Pignata2, Huey Jin Lim3, Sudha Suriyamurthy1, Kenneth T.Chang4,Victor K. Lee3, Ernesto Guccione2, and Reshma Taneja1

Abstract

Alveolar rhabdomyosarcoma (ARMS) is an aggressive suppressor PTEN was a direct target gene of G9a. G9a pediatric cancer with poor prognosis. As transient and stable repressed PTEN expression in a methyltransferase activity– modifications to chromatin have emerged as critical dependent manner, resulting in increased AKT and RAC1 mechanisms in oncogenic signaling, efforts to target epige- activity. Re-expression of constitutively active RAC1 in G9a- netic modifiers as a therapeutic strategy have accelerated in deficient tumor cells restored oncogenic phenotypes, demon- recent years. To identify chromatin modifiers that sustain strating its critical functions downstream of G9a. Collectively, tumor growth, we performed an epigenetic screen and our study provides evidence for a G9a-dependent epigenetic found that inhibition of lysine methyltransferase G9a sig- program that regulates tumor growth and suggests targeting nificantly affected the viability of ARMS cell lines. Targeting G9a as a therapeutic strategy in ARMS. expression or activity of G9a reduced cellular proliferation and motility in vitro and tumor growth in vivo.Transcrip- Significance: These findings demonstrate that RAC1 is an tome and chromatin immunoprecipitation–sequencing effector of G9a oncogenic functions and highlight the poten- analysis provided mechanistic evidence that the tumor- tial of G9a inhibitors in the treatment of ARMS.

Introduction the expression of PAX3-FOXO1 and PAX7-FOXO1 fusion proteins in 60% and 20% cases, respectively (2, 3). PAX3-FOXO1 is Rhabdomyosarcoma (RMS) is a highly malignant soft-tissue believed to be critical for oncogenesis by control of genes required sarcoma of childhood, which accounts for approximately 5% to for proliferation, survival, and metastasis (4–8). ARMS expressing 8% of pediatric cancers, occurring mostly in children below 10 either of these translocations is termed fusion positive, whereas a years of age. Despite expression of the myogenic proteins MyoD small percentage of ARMS tumors (20%) are devoid of these and myogenin, RMS cells fail to undergo myogenic differentiation chromosomal translocations and are termed fusion negative. and are morphologically similar to premature mesenchymal Fusion-negative ARMS is similar to ERMS in terms of prognosis, cells (1). The most common forms of RMS are embryonal rhab- event-free survival, frequency of metastases, and distribution of domyosarcoma (ERMS) and alveolar rhabdomyosarcoma site signifying the importance of PAX3/PAX7-FOXO1 in fusion- (ARMS). ARMS, the most aggressive subtype, is associated with positive ARMS tumors (9). In addition to the presence of fusion frequent metastasis at the time of diagnosis and exhibits limited oncoproteins, comprehensive analysis of the genomic landscape response to treatment, resulting in poor survival rates. ARMS is in ARMS has shown deregulation of growth factor signaling distinguished from the other subtypes of RMS by frequent trans- including insulin-like growth factor (IGF)/insulin-like growth locations t(2; 13) (q35; q14) and t(1; 13) (p36; q14), resulting in factor receptor 1, mutations in PIK3CA, and ampliﬁcation of CDK4 (10–12). Elevated growth factor levels stimulate the PI3K pathway, resulting in phosphorylation of phosphatidylinositol- 1Department of Physiology, Yong Loo Lin School of Medicine, National University 4,5-bisphosphate (PIP2) to produce phosphatidylinositol-3,4,5- of Singapore, Singapore, Singapore. 2Institute of Molecular and Cell Biology bisphosphate (PIP3). This facilitates recruitment of AKT to the (IMCB), Agency for Science, Technology and Research (ASTAR), Singapore, plasma membrane. Subsequent phosphorylation of AKT at thre- 3 Singapore. Department of Pathology, Yong Loo Lin School of Medicine, onine 308 (Thr308) by phosphoinositide-dependent kinase National University of Singapore, Singapore, Singapore. 4Department of Pathol- (PDK1) and at serine 473 (Ser473) by the mTORC-2 leads to its ogy, KK Women and Children's Hospital, Singapore, Singapore. activation and affects several cellular processes such as prolifer- Note: Supplementary data for this article are available at Cancer Research ation, growth, survival, migration, and metabolism (13). The Online (http://cancerres.aacrjournals.org/). guanine nucleotide exchange factors are also activated by PI3K/ Corresponding Author: Reshma Taneja, National University of Singapore, Block AKT signaling, resulting in increased Ras-related C3 botulinum MD9, 2 Medical Drive, Singapore 117593, Singapore. Phone: 65-6516-3236; toxin substrate 1 (RAC1) activity (14). Elevated RAC1 activity is Fax: 65-6778-8161; E-mail: [email protected] associated with aggressive cancers (15) and has multiple effects on doi: 10.1158/0008-5472.CAN-18-2676 growth, proliferation, and cell invasion. The tumor-suppressor 2019 American Association for Cancer Research. PTEN dephosphorylates PIP3 and thus negatively regulates the

2232 Cancer Res; 79(9) May 1, 2019

G9a Regulates RAC1 Activity in ARMS

PI3K–AKT pathway. The downregulation of PTEN expression in ma-Aldrich). For transient knockdown experiments, cells were ARMS contributes to elevated PI3K/AKT/RAC1 signaling that is transfected with 100 nmol/L of human G9a-specific siRNA (ON- linked to tumor cell survival (16, 17). However, the molecular TARGETplus EHMT2 siRNA SMARTpool, Dharmacon) contain- mechanisms that result in loss of PTEN expression are not under- ing a pool of three to five 19–25 nucleotide siRNAs. Control cells stood and could provide a tool to suppress oncogenic signaling were transfected with 100 nmol/L scrambled siRNA (ON-TAR- through the PI3K/AKT/RAC1 pathway. GETplus, nontargeting pool, Dharmacon) using Lipofectamine Recent advances in chromatin biology have revealed deregula- RNAiMax (Thermo Fisher Scientific). Cells were analyzed 48 tion of the epigenetic landscape in ARMS (18). The chromatin hours after transfection for all assays. For generating stable knock- modifiers CHD4, PCAF, and BRD4 have been shown to interact down cell lines, 293FT cells were transfected with packaging with and affect PAX3-FOXO1 activity (19–21). Nevertheless, the plasmids plP1 (5 mg) and plP2 (5 mg), envelope plasmid plP/ epigenetic network that is central to ARMS tumorigenesis remains VSV-G (5 mg; ViraPower Lentiviral Packaging Mix, Thermo Fisher to be identified. EHMT2/G9a and EHMT1/GLP are key SET domain Scientific), and 5 mg lentiviral expression constructs shRNA containing lysine methyltransferases that catalyze dimethylation of (pLKO.1, Mission shRNA DNA clone, Sigma-Aldrich) or shG9a histone H3 lysine 9 (H3K9me2), resulting in transcriptional repres- (pLKO.1) (#SHCLND-NM_025256 MISSION shRNA Plasmid sion of target genes (22). Deregulated expression and function of DNA). Sixteen hours after transfection, the cell supernatant was G9a, and to a lesser extent GLP, have been reported in cancers of replaced with basal DMEM. The cell supernatants were collected different origins (23–34). In addition to its canonical role in gene 48 and 72 hours after transfection, centrifuged, and the viral pellet repression, G9a can also activate genes through association with was resuspended in 200 mL of RPMI medium as stock solution. coactivator complexes in a context-dependent manner (23). How- RH30 cells at 40% to 50% confluency were transduced with ever, the signaling pathways that G9a regulates and its downstream shRNA control virus (control), or shG9a virus and 2 mL polybrene effectors are poorly understood. In the context of skeletal myogen- (8 mg/mL; Sigma-Aldrich) in RPMI 1640 basal medium. Six hours esis, we and others have shown that G9a inhibits muscle differen- after transduction, cell supernatants were replaced with RPMI tiation and promotes proliferation of myoblasts (35–38). We medium (10% FBS) for 24 hours. Transduced cells were selected therefore hypothesized that G9a expression may be deregulated with 1 mg/mL puromycin (Sigma-Aldrich) for 3 days. For rescue in ARMS and contribute to tumorigenesis. experiments, shcontrol and shG9a cells were transfected with 2 mg Here, we provide evidence that the G9a pathway is central to of Rac1-GFP plasmid (pcDNA3-EGFP-Rac1-Q61L, Addgene) using ARMS tumorigenesis. Using an epigenetic screen, we demonstrate Lipofectamine 2000 with PLUS reagent (Thermo Fisher Scientific). that G9a inhibitors potently affect viability of ARMS cell lines. Downregulation of G9a expression or pharmacologic inhibition Drug sensitivity of ARMS cells to inhibitors of of its activity not only reduces tumor cell proliferation and methyltransferases invasion, but also results in myogenic differentiation. In vivo, RH30 and RH4 cells were treated for 8 days with the indicated reduction of G9a function reduces tumor growth in xenograft compounds in 384 wells at three different concentrations (3, 1, mouse models. Transcriptomic and chromatin immunoprecipi- and 0.3 mmol/L). Viability at days 4 and 8 was scored by MTS assay tation (ChIP)-sequencing (ChIP-seq) analyses demonstrate and reported as the ratio over control-treated cells (with equiv- that G9a directly inhibits PTEN expression in a methylation- alent dilution of DMSO; RED > control; WHITE ¼ control; BLUE < dependent manner, resulting in the upregulation of phospho- control). The experiment was conducted in triplicate, and AKT levels and RAC1 activity. Re-expression of constitutively (þ)-JQ1 was used as a positive control. active RAC1 reverses the impact of G9a deficiency in tumor growth. Our data suggest that the PTEN–RAC1 axis downstream Western blot analysis of G9a is central to ARMS tumorigenesis. Whole-cell extracts were isolated using RIPA or SDS lysis buffer supplemented with protease inhibitors (Complete Mini, Sigma- Aldrich). The following primary antibodies were used: anti-G9a Materials and Methods (#3306S, 1:300, Cell Signaling Technology), anti-PTEN (138G6) Cell culture and stable cell lines (#9559, 1:1,000, Cell Signaling Technology), anti–phospho-AKT Primary human skeletal muscle myoblasts (HSMM; Lonza Inc.) (Ser473) (D9E) (#4060, 1:500, Cell Signaling Technology), were cultured in growth medium (SkGM-2 BulletKit). ARMS cell anti-AKT (#9272, 1:1,000, Cell Signaling Technology), anti- lines RH30, RH4, and RH41 were provided by Peter Houghton H3K9me2 (#9753S, 1:1,000, Cell Signaling Technology), anti- (Nationwide Children's Hospital, Columbus, OH) and Rosella RAC1 (23A8) (#05-389, 1:500, Merck Millipore), anti–cyclin Rota (Bambino Gesu Children's Hospital, Rome, Italy) and D1 (H295) (#sc753, 1:250, Santa Cruz Biotechnology), anti–ß- cultured in RPMI 1640 with L-glutamine (Thermo Fisher Scien- actin (#A2228, 1:10,000; Sigma-Aldrich), anti-H3 (#ab1791, tific) with 10% FBS (Hyclone). ARMS cells lines were authenti- 1:10,000; Abcam), and anti-GAPDH (14C10) (#2118, 1:500, cated for PAX3-FOXO1, MyoD, and myogenin expression by Santa Cruz Biotechnology). Appropriate secondary antibodies Western blot as described (21). Cell lines were tested at least (IgG-Fc–Specific Peroxidase) of mouse or rabbit origin (Sigma- every 6 months for Mycoplasma contamination using the Mycokit Aldrich) were used. Detection Kit from Biowest (K1000). After thawing, cell lines were not passaged more than 10 times. The cell lines were maintained RAC1 activity assay for no more than 3 passages between experiments. Where indi- RAC1 activity in control and shG9a cells and RH30 cells cated, cells were treated with 2.5 mmol/L UNC0642 (Sigma- untreated and treated with UNC0642 was detected using the Aldrich) or RAC1 inhibitor NSC23766 (Tocris Bioscience) at Active RAC1 Detection Kit (Cell Signaling Technology). GST- 50 mmol/L (RH30) and 30 mmol/L (RH41) for 48 hours. Control PAK1-PBD fusion protein was used to bind the activated form cells were treated with equivalent concentrations of DMSO (Sig- of GTP-bound RAC1. The complex was immunoprecipitated with

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2233

Bhat et al.

glutathione resin through GST-linked binding protein. Unbound For qPCR, total RNA was extracted using TRIzol (Thermo Fisher proteins were washed off during centrifugation, and glutathione Scientific) and quantified using Nanodrop. mRNA was converted resin-bound GTPase was eluted with SDS buffer. Active RAC1 to a single-stranded cDNA using iScript cDNA Synthesis Kit (Bio- levels were determined by Western blot using anti-RAC1 antibody Rad). qPCR was performed using Lightcycler 480 SYBR Green 1 (Cell Signaling Technology). Master Kit (Roche). PCR amplification was performed as follows: 95C for 5 minutes, followed by 95C for 10 seconds, annealing at Proliferation, migration, and invasion assays 60C for 10 seconds, followed by 45 cycles at 72C for 10 seconds. Cell proliferation was assessed by 5-bromo-20-deoxy-uridine Melting curves were generated and tested for a single product after (BrdU) labeling (Roche). Briefly, cells grown on coverslips were amplification. Light Cycler 480 software (version 1.3.0.0705) was pulsed with 10 mmol/L BrdU for 30 minutes, fixed, and stained used for analysis. CT values of samples were normalized to with anti-BrdU antibody (1:100) for 30 minutes. The coverslips internal control GAPDH to obtain delta CT (DCT). Relative D were washed, incubated with anti-mouse Ig-fluorescein antibody expression was calculated by 2 CT equation. qPCR was done (1:200), and mounted onto a glass slide using DAPI (Vectashield, using reaction triplicates, and at least two independent biological Vector Laboratories). Images were captured using fluorescence replicates were done for each analysis. Representative data are microscope BX53 (Olympus Corporation). shown. Error bars indicate the mean SD. Primer sequences for For transwell migration assays, cells were seeded in a 6-well PTEN, RAC1, and PI3KR1 have been described (39–41). Primers dish at a confluency of 1 105 cells/well. Where appropriate, cells for G9a are 50-TGGGCCATGCCACAAAGTC-30 and 50-CAGATG- were treated with UNC0642 (2.5 mmol/L) or siG9a (100 nmol/L) GAGGTGATTTTCCCG-30. The transcriptome data are compliant for 48 hours and serum starved for 24 hours. Note that 5 104 with MIAME guidelines and have been submitted to the GEO cells were seeded into extracellular matrix (Matrigel, collagen I)– repository (accession number GSE118777). coated polycarbonate membrane inserts (8.0 mm pore size) and placed in a 24-well plate containing RPMI 1640 (10% FBS). After ChIP 24 hours, cells that migrated to the bottom surface of the insert ChIP-seq was performed with anti-G9a antibody (Abcam). were fixed with 4% paraformaldehyde and stained with 6% crystal Twenty million RH41 cells were treated with protein cross-linker violet for 20 minutes. Cells were then counted based on five field disuccinimidyl glutarate, followed by chromatin cross-linking digital images taken at 10X magnification using EVOS XL core with 1% formaldehyde. Cells were lysed with Farnham lysis buffer imaging system AMEX1000 (Thermo Fisher Scientific). (5 mmol/L PIPES, pH 8.0, 85 mmol/L KCl, and 0.5% NP-40, with For wound-healing assays, two wounds were created perpen- protease inhibitors) and passed through 27-gauge needle. Crude dicular to each other using a 200 mL yellow tip in the plates with nuclear preparation was collected by centrifugation and lysed cells at approximately 95% confluency. The wound was then with RIPA buffer (1X PBS, 1% NP-40, 0.5% sodium deoxycholate, monitored at 24 hours. The migratory capacity was calculated as and 0.1% SDS with protease inhibitor). Chromatin shearing was a percentage of wound closure with respect to zero-time point. performed using Bioruptor (Diagenode). Washes were performed The cells were fixed and imaged as described above. using buffers provided in the ChIP Kit (#17-295, Merck Milli- pore). Chromatin complex was eluted with elution buffer, reverse Immunofluorescence assays cross-linked, and treated with proteinase K. DNA was extracted For differentiation assays, RH30 and RH41 cells were cultured using phenol-chloroform-isoamyl alcohol. Note that 30 ng for 3 to 5 days in RPMI 1640 supplemented with 2% horse ChIPed DNA and corresponding input were used for ChIP-seq serum (Gibco) at 90% to 95% confluency. Cells were fixed with library construction (New England Biolabs). Library fragment size 4% paraformaldehyde, and incubated with anti-Myosin Heavy was determined using the DNA 1000 Kit on the Agilent Bioana- Chain (MHC; R&D Systems; 1:500, 1 hour at room temperature), lyzer (Agilent Technologies). Libraries were quantified by qPCR followed by secondary goat anti-Mouse IgG (HþL) Highly Cross- using the KAPA Library Quantification Kit (KAPA Biosystems). Adsorbed Secondary Antibody, Alexa Fluor 568 (Thermo Fisher Libraries were pooled in equimolar, and sequencing (150 bp pair- Scientific). Coverslips were mounted with DAPI (Vectashield, end) was performed on the Illumina MiSeq at the Duke-NUS Vector Laboratories) and imaged using upright fluorescence Genome Biology Facility, according to the manufacturer's proto- microscope BX53 (Olympus Corporation). col (Illumina). The total analyzable reads were computed after filtering reads with low mapping score (MAPQ < 10) and PCR- Transcriptome analysis and quantitative real-time PCR duplicated reads. To evaluate immunoprecipitation, the fraction Multiplex gene expression analysis was carried out using Pan- of reads in peaks (FRiP) was computed using deeptools. The Cancer pathway Panel (#XT-CSO-PATH1-12, Nanostring) using library of G9a ChIP-seq had an FRiP score of 23%, indicating that RNA from two biological replicates of RH41 cells transfected with the library achieved successful global ChIP enrichment. Predic- scrambled siRNA or siG9a. RNA clean-up was performed using tion of G9a-binding sites was done using MACS2 to capture both RNeasy MiniElute Cleanup Kit (Qiagen). The hybridized RNA was broad (q value < 0.10) and narrow peaks (q value < 0.05). The scanned using nCounter digital analyzer. The RCC files generated prediction revealed 21,506 binding sites. The unique number of from nCounter were analyzed using nSolver 3.0 software. Raw predicted binding sites from chr1-chrX was used in the analysis. counts were normalized to the geometric mean of positive control The average fold enrichment of multiple peaks from the same and housekeeping genes. To identify genes and pathways that binding site was computed. To assess whether the distribution of are deregulated, PanCancer Pathway Advanced Analysis module G9a bindings at promoters, gene body, and intergenic regions is was applied. The PanCancer pathway data are compliant with not obtained from random chance, we compared the total num- Minimum Information About a Microarray Experiment (MIAME) ber of G9a bindings overlapping with each genomic feature guidelines and have been deposited in NCBI Gene Expression against a random distribution. Particularly, we shuffled the loca- Omnibus (GEO) Database (GEO accession number GSE118777). tion of G9a-binding sites and computed the total number of

2234 Cancer Res; 79(9) May 1, 2019 Cancer Research

G9a Regulates RAC1 Activity in ARMS

random bindings overlapping with each genomic feature. This tumors were fixed in paraformaldehyde. Paraffin sections were procedure was repeated for 10,000 iterations. Empirical P values stained with hematoxylin and eosin or analyzed by IHC. In an were then computed as the fraction of total iterations where the alternative approach, shcontrol and shG9a cells were injected in number of random bindings is greater than the observed number mice, and tumor growth was monitored as described above. For of G9a bindings overlapping with each feature. We detected IHC, slides were deparaffinized and rehydrated, followed by 21,506 G9a-binding sites across the genome, of which 44% antigen retrieval with Na-citrate buffer (pH 7.0). Sections were (n ¼ 9,560) localized at promoters, 31% (n ¼ 6,623) at gene incubated overnight at 4C with anti-G9a (1:200, Abcam), bodies, and 25% (n ¼ 5,323) at intergenic regions. Interestingly, anti-H3K9me2 (Abcam), anti-Ki67 (1:100, Santa Cruz Biotech- the overlap of G9a-binding sites at promoters was approximately nology), anti-active caspase 3 (Asp-175, 1:100; Cell Signaling 3x higher than the random bindings at promoters, and was Technology), anti-RAC1 (23A8; Merck Millipore), anti-active statistically significant (empirical P value < 0.0001), highlighting RAC1 (1:100, New East Biosciences) antibodies, followed by that the distribution of G9a binding at promoter was not obtained biotinylated goat anti-rabbit/anti-mouse IgG (HþL) secondary by random chance. On the other hand, the median occurrence of antibody (Vector Laboratories) for 1 hour at 37C. Sections were random binding events at gene bodies and intergenic regions were washed and incubated with Vectastain Avidin–Biotin Complex 9,020 and 9,425, respectively, indicating that the observed G9a- (Vector Laboratories) for 20 minutes at 37C. After incubation, binding sites at these genomic features were underrepresented DAB substrate (Vector Laboratories) was added until color compared with random events. Top 5,000 predicted G9a-binding development. Sections were counterstained with hematoxylin sites showing high fold enrichment were used for the association (Sigma-Aldrich). Slides were dehydrated and mounted using DPX analysis with biological processes. This was performed using (Sigma-Aldrich) and imaged using BX53 Olympus microscope. the Genomic Regions Enrichment of Annotations Tool (GREAT) IHC on 15 primary fusion-positive archival tumor tissues (http://great.stanford.edu/public/html/) with default parameters. obtained from the National University Hospital and KK Women's To ensure consistency of the results, different top predicted and Children Hospital in Singapore was done essentially as binding sites (n ¼ 2,500; n ¼ 10,000) were also tested. described (21). Sections were stained with anti-G9a antibody Sequencing reads mapped against hg19- and G9a-bound (1:50 dilution, Cell Signaling Technology). Three normal muscle regions were identified by MACS2000-predicted binding sites tissue sections were used as controls. Negative controls were per- showing high fold enrichment (binding signal to background formed using secondary antibody only. Images were captured with ratio). The observation remained consistent even by using a Olympus BX43 microscope. Approval for the study was obtained different number of predicted sites (n ¼ 2,500; n ¼ 10,000), from the ethics committee (Institutional Review Board) at NUS. suggesting that the associations with biological processes are robust and independent from the arbitrary number of sites. The Statistical analysis ChIP-seq data are compliant with MIAME guidelines and have Error bars indicate mean SD unless specified otherwise. been deposited in the NCBI GEO database (accession number Significance was determined using the Student t test (two-sided), GSE118666). ChIP-PCR was done as previously described (38). and P values less than 0.05 were considered statistically significant Briefly, 5 106 cells were cross-linked with 1% formaldehyde (, P < 0.05; , P < 0.01; and , P < 0.001). for 10 minutes at 37C, sonicated, and ChIP was carried out according to the kit protocol (Merck Millipore). Immunopre- Data availability cipitates were reverse cross-linked, DNA extracted, and qPCR The ChIP-seq data have been deposited in GEO under the performed as described above. Ten percent of the initial DNA accession number GSE118666. The PanCancer pathway data have was used as input control. Relative enrichment was calculated been deposited in GEO under the accession number GSE118777. D using 2 CT equation. The following antibodies were used for ChIP assays: ChIP-grade anti-G9a (Abcam), anti-H3K9me2 (Abcam), anti-H3K9ac (Abcam), anti-H3K27ac (Abcam). Pri- Results mers sequences for CSF2 and PGC1a have been described (42). G9a is an overexpressed and druggable target in ARMS G9a occupancy at the PTEN was analyzed using the primers: We performed a small-molecule screen in order to identify Forward: 50-GCAGGAAGGGTTGGGGTTCC-30 and Reverse: 50- druggable methyltransferases involved in the tumorigenesis of GGATACACGGGCCACAGTCG-30 as described (42). ARMS. Three concentrations (3, 1, and 0.3 mmol/L) of 13 methyltransferases inhibitors (provided by the Structural Genomic Con- Mouse xenograft experiments and IHC sortium) were tested in two different cell lines, RH30 and RH4, at All animal procedures were approved by the Institutional two different time points (days 4–8). Viability was evaluated via an Animal Care and Use Committee. Note that 6 106 RH30 cells MTS assay (Fig. 1A). Blue and red in the heatmap indicate a lower or were injected subcutaneously into the right flank of 6-week-old higher number of cells, respectively, compared with the control. In CrTac:NCr-Foxn1 nude female mice (InVivos) as described (21). addition to JQ1 that inhibits Brd4 activity and has recently been When tumor nodules were palpable, mice (n ¼ 10/group) were shown to affect PAX3-FOXO1 in RMS (19), the screen identified injected intraperitoneally with UNC0642 at 5 mg/kg body weight four drugs that show a strong effect on viability in both cell lines: every alternate day. Control group was injected with DMSO. MS023 (PRMT type I inhibitor), GSK591 (PRMT5 inhibitor), Tumor diameter was measured using digital vernier caliper on UNC0638, and UNC0642 (G9a inhibitors). We decided in this alternate days. Tumor volume was calculated using the following study to focus on the oncogenic activity of G9a. The effect of formula: V ¼ (L W W)/2, where V is the tumor volume, W is modulating arginine methylation in ARMS will be described else- the tumor width, and L is the tumor length. Body weight was where. Consistent with the screen, treatment of patient-derived measured every alternate day. Once the tumors reached 1.5 cm in RH30 and RH41 cells with UNC0642 (Supplementary Fig. S1A diameter in the control cohort, mice were euthanized and resected and S1B), or downregulation of G9a expression using shRNA

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2235

Bhat et al.

Color key A RH30 RH4 Color key RH30 RH4 Alveolar Alveolar Day 4 Alveolar Alveolar Day 8 0 2 (Pax3-Foxo1) (Pax3-Foxo1) (Pax3-Foxo1) (Pax3-Foxo1) DRUGS TARGETS 0 2 Value m m m m m m DRUGS TARGETS 3 mol/L 1 mol/L 0.3 mol/L 3 mol/L 1 mol/L 0.3 mol/L Value 3 mmol/L 1 mmol/L 0.3 mmol/L 3 mmol/L 1 mmol/L 0.3 mmol/L GSK343 EZH2, EZH1 BAY-598 SMYD2

SGC707 PRMT3 DMSO

R)-PFI-2 R)-PFI-2 SETD7 (hydrochloride) (hydrochloride) SETD7

DMSO SGC0946 Dot1L

BAY-598 SMYD2 SGC707 PRMT3

A-366 G9a, EHMT1 GSK343 EZH2, EZH1

SGC0946 Dot1L A-366 G9a, EHMT1

UNC1999 EZH2, EZH1 MS049 (HCl salt) PRMT4/6

A-196 SUV420H1/2 A-196 SUV420H1/2

MS049 (HCl salt) PRMT4/6 MS023 (HCl salt) PRMT type I

MS023 (HCl salt) PRMT type I UNC0642 G9a, EHMT1

UNC0642 G9a, EHMT1 GSK591 PRMT5

UNC0638 G9a, EHMT1 UNC1999 EZH2, EZH1

GSK591 PRMT5 UNC0638 G9a, EHMT1

(+)-JQ1 BRD4 (+)-JQ1 BRD4 B *** C 12 *** DRUGS RH30 RH4 10 MS023 300 nmol/L 2,500 nmol/L 8 UNC0642 300 nmol/L 2,000 nmol/L 6 G9a - 160 4 UNC0638 2,000 nmol/L 2,000 nmol/L 2 GSK591 <300 nmol/L 750 nmol/L b-Actin -42

Relative expression 0 D Negative Normal control muscle ARMS patient tumors (PAX3-FOXO1 positive )

1 1 1 4 7 10 13

2 2 2 5 8 11 14

333 6 9 12 15

Figure 1. G9a is overexpressed in fusion-positive ARMS. A, Drug sensitivity of ARMS cell lines RH30 and RH4 to methyltransferase inhibitors was tested. Cell lines were treated up to 8 days with the indicated compounds in 384 wells (3, 1, and 0.3 mmol/L). Viability at days 4 and 8 was scored by MTS assay and reported as the ratio over control-treated cells (with equivalent dilution of DMSO; RED > control; WHITE ¼ control; BLUE < control). The experiment was conducted in triplicates, and (þ)-JQ1 was used as a positive control. The results showed that MS023, UNC0642, UNC638, and GSK591 had a strong effect on the viability of ARMS cell lines.

The IC50 of MS023, UNC0642, UNC0638, and GSK591 in RH30 and RH4 after 8 days is shown. B and C, G9a mRNA (B) and protein (C) levels were examined in primary HSMM and ARMS cells lines (RH30 and RH41) by qPCR and Western blotting, respectively. Error bars, mean of qPCR triplicates in each experiment SD. The results are representative of two independent experiments. Numbers indicate the molecular weight of proteins. D, G9a expression in three normal skeletal muscles (Normal 1–3) was compared with 15 archival ARMS patient tumor specimens (#1–15) by IHC using anti-G9a antibody. Higher nuclear G9a staining was apparent in all tumors, albeit to varying levels. The upper right inset shows a zoomed-in image. Negative controls were done by immunostaining three tumor specimens with secondary antibody alone. Scale bar, 50 mm. , P < 0.001.

2236 Cancer Res; 79(9) May 1, 2019 Cancer Research

G9a Regulates RAC1 Activity in ARMS

A RH30 DAPI MHC Merge * 60

40 Control

% BrdU positivity 0 Control shG9a shG9a

B C RH30 1 DAPI MHC Merge RH30 *** 6 0.8 * 0.6

* Control 0.4

0.2 Cell density *10 Cell density 0 D0 D1 D2 D3 D4 D5 shG9a sh control shG9a D5 D E RH30 RH30 0 hr 24 hr * *** 100 100 80 80 Control

Control 60 60 40 40 % Invasion %

% Migration 20 20 shG9a 0 0 shG9a Control shG9a Control shG9a 24 hr

Figure 2. Stable knockdown of G9a alters proliferation, differentiation, and motility. A, BrdU assays were done with control and shG9a cells. The percentage of proliferating cells was quantiﬁed. B, Growth curve assay was performed by seeding 0.1 million control and shG9a RH30 cells. Cells were trypsinized every 24 hours, and cell numbers were counted for a period of 5 days. C, Control and shG9a cells were differentiated for 5 days in 2% horse serum-containing medium. Cells were immunostained with anti-MHC antibody (red). D and E, Motility of control and shG9a cells was analyzed by wound-healing assays to assess migration (D) and by Matrigel Boyden chamber assays (E). Migration of cells was monitored over a 24-hour period. Error bars, mean SD. , P < 0.05; , P < 0.001.

(Supplementary Fig. S1C), led to a striking reduction in colony plementary Fig. S1C), or transiently, using siRNA in RH30 (Sup- formation (Supplementary Fig. S1D). These ﬁndings led us to plementary Fig. S2A) cell line. In addition, we inhibited its examine G9a expression in ARMS cell lines and tumor specimens. catalytic activity using UNC0642 (UNC) for 48 hours, which G9a mRNA and protein levels were upregulated in RH30 and RH41 resulted in reduced H3K9me2, a signature of G9a activity (Sup- cell lines (Fig. 1B and C) relative to primary HSMM. In addition, plementary Fig. S1A and S1B). A striking reduction in cellular IHC analysis of 16 archival PAX3-FOXO1 fusion-positive ARMS proliferation in shG9a cells was apparent by BrdU assay (Fig. 2A) tumor sections showed that G9a was overexpressed in all specimens that was conﬁrmed by growth curve assays (Fig. 2B). A similar in comparison with normal human muscle (Fig. 1D). impact on cellular proliferation was seen in RH30 cells upon transient G9a knockdown (siG9a) and UNC treatment compared Knockdown of G9a expression or activity suppresses with their respective controls (Supplementary Fig. S2B). Interest- proliferation and migration, and enhances myogenic ingly, an increased propensity to undergo myogenic differentia- differentiation tion was also apparent by immunostaining with MHC antibody in To examine its function in ARMS, we used two approaches. We shG9a cells (Fig. 2C). These results were validated in siG9a RH30, down regulated endogenous G9a expression either stably (Sup- as well as in response to UNC treatment (Supplementary Fig.

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2237

Bhat et al.

S2C). Consistently, expression of myogenic markers was elevated with previous reports (42), we identifiedG9aoccupancyatthe in siG9a and UNC-treated RH30 cells (Supplementary Fig. S2D). PGC1a and CSF2 promoters by ChIP-seq that was validated by Because G9a has been reported to influence migratory and inva- ChIP-PCR (Supplementary Fig. S5B and S5C). sive capacity of cancer cells (22–24), we tested its involvement in Given the repression of its expression by G9a, we examined migration by wound-healing assays. Wound closure was moni- levels of the PI3K–AKT pathway that is regulated by PTEN. Upon tored for 24 hours in G9a knockdown cells or cells that had been UNC0642 treatment, as well as in siG9a cells, reduced phospho- pretreated with UNC0642 for 48 hours. Defective migration was AKT levels were apparent, indicating G9a regulates the PTEN– observed in shG9a cells (Fig. 2D), RH30 siG9a cells (Supplemen- pAKT axis in a methyltransferase activity–dependent manner tary Fig. S2E) as well as upon treatment with UNC0642 compared (Fig. 4A). In addition, a clear downregulation of active RAC1 with respective control cells. In addition, significantly fewer G9a levels was seen upon UNC0642 treatment as well as in shG9a cells knockdown cells and UNC0642-treated cells invaded through (Fig. 4B and C). Matrigel relative to controls (Fig. 2E; Supplementary Fig. S2F). We further validated the effects of G9a depletion in a second PAX3- Re-expression of active RAC1 restores proliferative and þ FOXO1 ARMS cell line RH41. Both siG9a (Supplementary Fig. metastatic capacity of shG9a cells S3A) and UNC-treated cells showed reduced proliferation relative Because RAC1 activity was down regulated by loss of G9a, we to their controls (Supplementary Fig. S3B). Myogenic differenti- tested whether it is an effector of G9a. We first determined whether ation was increased as seen by MHC staining and elevated Myh1 inhibition of RAC1 activity mimics G9a deficiency in ARMS cell expression (Supplementary Fig. S3C and S3D) in siG9a and UNC- lines. We used NSC23766, a small-molecule inhibitor of active treated cells, whereas cellular migration and invasion were RAC1. Cells treated with NSC23766 showed marked reduction of þ decreased (Supplementary Fig. S3E and S3F). BrdU cells (Supplementary Fig. S6A). In addition, migration and invasion were significantly reduced in cells treated with PI3K signaling pathway is differentially regulated in ARMS cells NSC23766 compared with controls (Supplementary Fig. S6B and silenced for G9a S6C). We then examined if re-expression of constitutively active To identify the molecular pathways and functional targets RAC1 restores proliferation and motility defects observed in downstream of G9a, we performed Nanostring PanCancer path- shG9a cells. To this end, we transfected control and shG9a cells way analysis using RNA from control and siG9a cells. Global with constitutively active RAC1 (RAC1Q61L) or empty vector significance scores indicated that PI3K signaling was among the (Fig. 4D). The proliferative capacity of shG9a cells was signifi- most differentially regulated pathways (Fig. 3A; Supplementary cantly rescued upon expression of RAC1Q61L-GFP compared Table S1A). Of the 55 genes that were significantly altered by loss with empty vector control (Fig. 4E). In addition, there was of G9a expression, several genes in PI3K signaling pathway were significant rescue of motility in shG9a cells as seen by wound- found to be deregulated (Fig. 3B; Supplementary Table S1B). To healing and invasion assays (Fig. 4F and G). validate these findings, we focused on PTEN, an upstream regulator of the PI3K/AKT pathway. Consistent with the tran- G9a knockdown inhibits tumorigenicity in vivo scriptome analysis, PTEN mRNA was significantly upregulated in To validate the role of G9a in tumorigenesis in vivo, we gener- G9a knockdown cells, as well as upon UNC0642 treatment in both ated xenograft models by injecting RH30 cells subcutaneously in cell lines compared with respective controls (Fig. 3C and D), nude mice. We used two approaches. In the first, mice were treated indicating that it is regulated by G9a in a methylation-dependent with UNC0642 once tumors were palpable. The control group manner. The expression of additional differentially expressed genes was treated with DMSO. Tumor volumes and body weights were was validated by qPCR in siG9a cells (Supplementary Fig. S4). measured. In the UNC0642-treated cohort, tumor volume was significantly lower compared with the matched DMSO controls G9a directly regulates PTEN expression and RAC1 activity (Fig. 5A–C) with no significant change in body weight (Fig. 5D). Because genome-wide direct targets of G9a in RMS are Tumor sections from DMSO control and UNC0642-treated unknown, we performed ChIP, followed by high-throughput groups were analyzed histologically and by IHC. A significant sequencing with a G9a-specific antibody. G9a occupancy was reduction in Ki67 staining was seen, indicating reduced tumor cell enriched at promoters, gene bodies, and intergenic regions proliferation. In addition, H3K9me2 staining was expectedly (Fig. 3E), and 44% of the peaks were mapped to promoters reduced in the UNC0642-treated mice compared with controls at the transcription start site (TSS; Fig. 3F). Gene ontology (Fig. 5E). No significant changes were seen in the staining for G9a analysis using GREAT showed that genes with G9a peaks were and cleaved caspase 3. Interestingly, active RAC1 levels were significantly enriched for biological processes that include strongly down regulated correlating with the in vitro data. More- skeletal muscle development, regulation of stem cell, or mes- over, a significant increase in PTEN expression was observed in the enchymal cell proliferation (Fig. 3G). G9a enrichment was tumors (Fig. 5F). In the second approach, we subcutaneously evident at the PTEN promoter region by ChIP-seq (Fig. 3H), injected control and shG9a cells in nude mice. Tumor volume was and its occupancy was validated by ChIP-PCR in RH30 and significantly lesser in mice injected with shG9a cells compared RH41 cells (Fig. 3I; Supplementary Fig. S5A). We then analyzed with controls (Fig. 5G–I), with no overt change in body weight activation (H3K9ac) and repression (H3K9me2) marks at the (Fig. 5J). Ki67 and active RAC1 levels were strikingly reduced, PTEN promoter by ChIP-PCR in control and shG9a cells, as whereas PTEN expression was elevated in the tumors (Fig. 5L). well as upon UNC0642 treatment (Fig. 3J and K). Consistent with an upregulation of its expression, H3K9ac was increased, whereas H3K9me2, a signature of G9a activity, was expectedly Discussion reduced. A similar trend with both histone marks was seen at G9a has been reported to be overexpressed in a variety of the PTEN promoter in response to UNC0642 treatment. In line cancers (24–34), yet the signaling pathways that it controls and

2238 Cancer Res; 79(9) May 1, 2019 Cancer Research

G9a Regulates RAC1 Activity in ARMS

A B RAC1 8 PTEN PIK3R1 6 AKT2 FAS 4 RELA 2 PIK3R4 AKT3 0 MMP9 expressed genes GNG4

Number of differentially 60 12 18 24 30 - 10 × log10 (P value)

C RH30 (PTEN mRNA) D RH41 (PTEN mRNA)

4 * 3 * 2.5 * 5 * 3 2 4 2 1.5 3 2 1 2 1 1 0.5 1 0 0 0 0 Relative expression Relative expression Relative expression Relative expression

Control Control

Biological processes E F G Skeletal muscle tissue development 100 Skeletal muscle organ development 25% Positive regulation of stem cell proliferation Regulation of stem cell proliferation 5.323 44% 80 Regulation of mesenchymal cell proliferation 9.560 60 Positive regulation of mesenchymal cell proliferation 31% Regulation of epithelial cell proliferation involved in lung morphogenesis 40 Positive regulation of multicellular organism growth 6.623 Neuron recognition

read density Positive regulation of epithelial cell proliferation involved in lung morphogenesis G9a CHIP-seq 20 Promoter 5 10 15 20 25 30 Gene body 0 log P value (Binomial test) 10 Intergenic -20 Kb TSS +20 Kb H I RH30 (PTEN promoter) * 50 Kb 0.06

KLLN 0.05 ATAD1 PTEN 0.04 CFL1P1 70 0.03

% Input 0.02 0.01 (q-value = 0.07) 0

J RH30 (PTEN promoter) K RH30 (PTEN promoter) H3K9ac H3K9me2 H3K9ac H3K9me2 * ** ** *** 0.04 0.3 0.04 0.3 0.25 0.03 0.03 0.2 0.2 0.02 0.02 0.15 % Input % % Input 0.1 0.1 % Input 0.01 0.01 % Input 0.05 0 0 0 0

Control Control

Figure 3. PTEN is a direct G9a target gene. A, Nanostring PanCancer Pathway analysis was performed with two biological replicates of control and siG9a cells. Gene set enrichment analysis revealed that PI3K signaling was among the most significantly altered pathways with deregulated expression of a large number of genes. B, A list of genes in the PI3K–AKT pathway that were significantly altered in siG9a cells relative to controls. The significance levels are shown. C and D, PTEN mRNA levels in control and siG9a cells, as well as upon UNC treatment, were analyzed by qPCR in RH30 and RH41 cells. The results are representative of two independent experiments. Error bars, mean of qPCR triplicates in each experiment SD. E, Genome-wide G9a occupancy was analyzed by ChIP-seq. A pie chart showing genomic distribution of G9a peaks at promoters, gene body, and intergenic regions. Highest G9a occupancy was observed at the promoter regions, followed by gene body and intergenic regions. F, Histogram showing G9a distribution at the TSS 20 kb. G, MsigDB analysis using GREAT showed biological processes that are significantly regulated by G9a. H, Genomic snap shot using UCSC Genome Browser showing G9a enrichment at the PTEN promoter. I, ChIP- PCR validation showed G9a occupancy at the PTEN promoter. The results are representative of two independent experiments. Error bars, mean of qPCR triplicates in each experiment SD. J and K, ChIP-PCR analysis in control and shG9a cells for H3K9ac and H3K9me2 marks at the PTEN promoter. In addition, ChIP-PCR analysis was done in DMSO control and UNC0642-treated cells for H3K9ac and H3K9me2 enrichment. Error bars, mean of qPCR triplicates in each experiment SD. The results are representative of two independent experiments. , P < 0.05; , P < 0.01; , P < 0.001.

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2239

Bhat et al.

ACRH41 B RH41 RH30 DMSO UNC Control shG9a Control Active G9a -160 -21 RAC1 G9a -160 Active -21 Total -21 RAC1 RAC1 PTEN -54 Total -21 RAC1 GAPDH -36 p-AKT -60 (Ser473) GAPDH -36 β-Actin -42

RH30 D RH30 E Control shG9a Control shG9a Control +RAC1 shG9a +RAC1 Rac1-GFP _ _ + + _ _ ++ G9a - 160 DAPI

GFP -27

GAPDH -36 BrdU

F RH30 * Control Control+RAC1 250 ** ** 75 200 * ** *** 60 150 cells 45 + 100 30 shG9a shG9a+RAC1

% Invasion 50 15 % BrdU 0 0

Control Control ControlControl 24 hr

G RH30 Control Control + RAC1 shG9a shG9a + RAC1 250 *** *** 200 0 hr 150 100

% Migration % 50

24 hr 0

Control Control

Figure 4. RAC1 is an effector of G9a-dependent oncogenic phenotypes. A, PTEN expression and p-AKT (Ser473) levels were analyzed by Western blot upon treatment with 2.5 mmol/L UNC0642 for 48 hours, as well as in control and siG9a cells. The results are representative of two independent experiments. B, RAC1 activity was analyzed 48 hours after UNC0642 treatment by immunoprecipitation with anti-active RAC1 antibody. The blots were immunoblotted with total RAC1 antibody. The results are representative of two independent experiments. C, RAC1 activity was examined in control and shG9a cells. Total RAC1 antibody was used as the loading control. The results are representative of two independent experiments. D–G, Control and shG9a cells were transfected with constitutively active RAC1-GFP plasmid. Expression of G9a and RAC1 was analyzed by Western blot using anti-G9a and anti-GFP antibodies (D). BrdU assay was done to measure proliferation (E), Matrigel transwell assay was done for invasion (F), and wound-healing assay for migration (G). Quantiﬁcation of data in each assay is shown in the respective bar graphs. The results are representative of at least two independent experiments. Error bars, mean SD. , P < 0.05; , P < 0.01; , P < 0.001.

its transcriptional targets remain to be identiﬁed in different direct G9a target gene. We show G9a occupancy and its signature tumor types. In this study, we report an unprecedented role of H3K9me2 repressive marks at the PTEN promoter. Loss of G9a G9a in epigenetic regulation of PI3K pathway in ARMS. Through expression or pharmacologic inhibition of its methyltransferase transcriptome and ChIP-seq analysis, we identiﬁed PTEN to be a activity relieved repression of PTEN expression, resulting in

2240 Cancer Res; 79(9) May 1, 2019 Cancer Research

G9a Regulates RAC1 Activity in ARMS

A B C 5,000 DMSO ** 4,000 * 3,000 UNC0642 * 2,000 (5 mg/kg) DMSO volume *

DMSO 1,000 Relative tumor 0 D10D8D6D4D2D0 UNC UNC E DMSO UNC M1 M2 M3 Tumor 1 Tumor 2 Tumor 1 Tumor 2

D 40 H&E 30 DMSO UNC0642 (5 mg/kg) 20

10 Ki67

Body wt. (g) 0 D10D8D6D4D2D0 Casp3 F DMSO UNC DMSO UNC

PTEN -54 G9a

RAC1 -21 me2 β-Actin -42 H3K9 M1 M2 RAC1 G H Active Total RAC1 50 mmol/L Control I 8,000 Control * Control 6,000 shG9a 4,000 *

shG9a * volume shG9a 2,000 M1 M2 M3 Relative tumor 0 D12D10D8D6D4D2D0

J 50 Control shG9a K Control shG9a 40 Tumor 1 Tumor 2 Tumor 1 Tumor 2 30 20

10 H&E Body wt. (g) 0 D8D6D4D2D0 D12D10 L Ki67 Control Control

G9a - 160 G9a

PTEN -54 RAC1 RAC1 -21 Active β-Actin -42 Total RAC1 M1 M2 50 mmol/L

Figure 5. G9a inhibition impairs tumor growth in vivo. A and B, Nude mice (n ¼ 10/group) were injected with RH30 cells. When tumors were palpable, mice were treated with either vehicle DMSO or UNC0642 (5 mg/kg body weight). Three representative control and UNC0642-treated mice (M1, M2, and M3) and resected tumors are shown in A and B, respectively. C, Relative tumor volume in control and UNC0642-treated mice was measured every 2 days after DMSO or UNC0642 treatment. Day 0 (D0) refers to the day injections with DMSO or UNC0642 were started. D, Body weight of control and UNC0642-treated mice was measured every 2 days after treatment. E, Tumor sections from two independent DMSO and UNC0642-treated mice were stained with hematoxylin and eosin (H&E). Sections from the tumors were immunostained with anti-Ki67, anti-caspase 3 (Casp3), anti-G9a, anti-H3K9me2, and anti-active RAC1 and total RAC1 antibodies. F, Tumors were analyzed for PTEN and total RAC1 levels by Western blot. G and H, Control and shG9a RH30 cells were injected in nude mice (n ¼ 5/group). Mice were sacriﬁced at D12 after injection of cells. Representative images of mice and resected tumors are shown in G and H, respectively. I and J, Relative tumor volume and body weight of control and shG9a mice were measured every 2 days. K, Tumor sections from two independent DMSO- and UNC0642-treated mice were stained with hematoxylin and eosin. Sections were immunostained with anti-Ki67, anti-G9a, anti-active RAC1, and anti-RAC1 antibodies. L, Tumor lysates from control and shG9a mice were analyzed by Western blot with anti-PTEN, anti-RAC1, and anti–b-actin antibodies.

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2241

Bhat et al.

reduced AKT and RAC1 activity, and consequently impaired for therapeutic intervention (49, 50). Although a few tyrosine proliferation and migration of tumor cells. Interestingly, inhibi- kinase inhibitors have now been approved by the FDA, develop- tion of G9a also resulted in an increased propensity of tumor cells ment of drug resistance remains a major challenge with PI3K to undergo differentiation. Our results demonstrate that G9a not inhibitors. AKT inhibitors have also met with limited clinical only sustains oncogenic signaling in the tumors, but also blocks success. Similarly, no clinically effective drugs targeting RAC1 their ability to undergo myogenic differentiation. Our data are activity are available to date (51). Recent bioinformatics analyses consistent with a recent study that showed G9a-mediated repres- of cancer genome databases implicate several methyltransferases sion of PTEN expression in non–small cell lung cancer (43). and acetyltransferases in various human cancers including Previous studies have shown that the PI3K/AKT signaling is RMS (18, 52). In particular, misregulation of lysine methylation elevated in ARMS. This increase stems from both elevated growth has been reported in various cancers that make proteins that factors such as IGF-II and FGFR4, which are targets of PAX3- govern this modification attractive drug targets. Our studies FOXO1, and decreased PTEN levels (16, 17). In addition to suggest that the use of selective G9a inhibitors to increase PTEN functioning as an antagonist of the PI3K/AKT pathway, PTEN expression and dampen AKT/RAC1 activity could provide a novel also dephosphorylates other substrates such as p125FAK (44). therapeutic approach in ARMS. However, increased motility of PTEN / fibroblasts occurs due to increased AKT and RAC1 activity without changes in p125FAK Disclosure of Potential Conflicts of Interest levels, indicating that RAC1 mediates the effects PTEN in cell No potential conflicts of interest were disclosed. migration (45). Our data are in line with these findings. PTEN expression is derepressed by UNC0642, which correlates with Authors' Contributions reduced RAC1 activity. Several positive and negative regulators of Conception and design: A.V. Bhat, R. Taneja PTEN expression have been described (46). Among these, the zinc Development of methodology: A.V. Bhat, M.P. Kala, L. Pignata, H.J. Lim Acquisition of data (provided animals, acquired and managed patients, finger transcription factors Snail and Slug have been shown to provided facilities, etc.): V.K. Rao, S. Suriyamurthy, K.T. Chang, V.K. Lee repress PTEN expression. Interestingly, G9a has been found to Analysis and interpretation of data (e.g., statistical analysis, biostatistics, interact with Snail (25) to mediate epithelial-to-mesenchymal computational analysis): V.K. Rao, S. Suriyamurthy, R. Taneja transition in breast cancer. These studies together with our data Writing, review, and/or revision of the manuscript: R. Taneja suggest that Snail may recruit G9a to the PTEN promoter. We note Administrative, technical, or material support (i.e., reporting or organizing that RAC1 mRNA is reduced in G9a knockdown cells, suggesting data, constructing databases): A.V. Bhat, V.K. Rao, K.T. Chang Study supervision: E. Guccione, R. Taneja that G9a may affect RAC1 activity through additional mechanisms. Similar to G9a, RAC1 is a key regulator of cell migration, Acknowledgments – impairs cell-cycle exit, and inhibits myogenic differentiation (35 We thank Peter Houghton and Rosella Rota for ARMS cell lines, and Ooi Wen 38, 47, 48). In addition, active RAC1 levels are elevated in Fong, Genome Institute of Singapore, for bioinformatics analysis. ARMS (48). We show that an RAC1 inhibitor mimics the effects This work was supported by a National Medical Research Council grant of a G9a inhibitor, suggesting that G9a and RAC1 downregulation (NMRC/CBRG/0063/2014) to R. Taneja and E. Guccione. A.V. Bhat is supported has similar phenotypic outcomes. Because re-expression of con- by the President's Graduate Scholarship at the National University of Singapore. stitutively active RAC1 rescues the effects of G9a deficiency, our data demonstrate that RAC1 is a downstream effector of G9a The costs of publication of this article were defrayed in part by the payment of rather than functioning in a parallel pathway to regulate page charges. This article must therefore be hereby marked advertisement in tumorigenesis. accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The PI3K/AKT signaling is hyperactivated in many cancers and contributes to survival and growth of tumor cells. Therefore, Received August 25, 2018; revised December 17, 2018; accepted February 26, targeting this signaling axis provides an attractive opportunity 2019; published first March 4, 2019.

References 1. Keller C, Guttridge DC. Mechanisms of impaired differentiation in rhab- a biologically relevant target. Genes Chromosomes Cancer 2008;47: domyosarcoma. FEBS J 2013;280:4323–34. 510–20. 2. Linardic CM. PAX3-FOXO1 fusion gene in rhabdomyosarcoma. 8. Bernasconi M, Remppis A, Fredericks WJ, Rauscher FJ, Sch€afer BW. Induc- Cancer Lett 2008;270:10–8. tion of apoptosis in rhabdomyosarcoma cells through down-regulation of 3. Olanich ME, Barr FG. A call to ARMS: targeting the PAX3-FOXO1 gene in PAX proteins. Proc Natl Acad Sci U S A 1996;93:13164–9. alveolar rhabdomyosarcoma. Expert Opin Ther Targets 2013;17:607–23. 9. Williamson D, Missiaglia E, de Reynies A, Pierron G, Thuille B, Palenzuela 4. Wachtel M, Sch€afer BW. PAX3-FOXO1: zooming in on an "undruggable" G, et al. Fusion gene-negative alveolar rhabdomyosarcoma is clinically and target. Semin Cancer Biol 2018;50:115–23. molecularly indistinguishable from embryonal rhabdomyosarcoma. J Clin 5. Cao L, Yu Y, Bilke S, Walker RL, Mayeenuddin LH, Azorsa DO, et al. Oncol 2010;28:2151–8. Genome-wide identification of PAX3-FKHR binding sites in rhabdomyo- 10. Shern JF, Chen L, Chmielecki J, Wei JS, Patidar R, Rosenberg M, et al. sarcoma reveals candidate target genes important for development and Comprehensive genomic analysis of rhabdomyosarcoma reveals a land- cancer. Cancer Res 2010;70:6497–508. scape of alterations affecting a common genetic axis in fusion-positive and 6. Davicioni E, Finckenstein FG, Shahbazian V, Buckley JD, Triche TJ, Ander- fusion-negative tumors. Cancer Discov 2014;4:216–31. son MJ. Identification of a PAX-FKHR gene expression signature that 11. Seki M, Nishimura R, Yoshida K, Shimamura T, Shiraishi Y, Sato Y, et al. defines molecular classes and determines the prognosis of alveolar rhab- Integrated genetic and epigenetic analysis defines novel molecular sub- domyosarcomas. Cancer Res 2006;66:6936–46. groups in rhabdomyosarcoma. Nat Commun 2015;6:7557. 7. Mercado G E, Xia SJ, Zhang C, Ahn EH, Gustafson DM, Lae M, et al. 12. Anderson J, Gordon A, Pritchard-Jones K, Shipley J. Genes, chromo- Identification of PAX3-FKHR-regulated genes differentially expressed somes, and rhabdomyosarcoma. Genes Chromosomes Cancer 1999;26: between alveolar and embryonal rhabdomyosarcoma: focus on MYCN as 275–85.

2242 Cancer Res; 79(9) May 1, 2019 Cancer Research

G9a Regulates RAC1 Activity in ARMS

13. Hemmings BA, Restuccia DF. PI3K-PKB/Akt pathway. Cold Spring Harb 33. Wang YF, Zhang J, Su Y, Shen YY, Jiang DX, Hou YY, et al. G9a regulates Perspect Biol 2012;4:a011189. breast cancer growth by modulating iron homeostasis through the repres- 14. Han J, Luby-Phelps K, Das B, Shu X, Xia Y, Mosteller RD, et al. Role of sion of ferroxidase hephaestin. Nat Commun 2017;8:274. substrates and products of PI 3-kinase in regulating activation of Rac- 34. Casciello F, Al-Ejeh F, Kelly G, Brennan DJ, Ngiow SF, Young A, et al. G9a related guanosine triphosphatases by Vav. Science 1998;279:558–60. drives hypoxia-mediated gene repression for breast cancer cell survival and 15. Kazanietz MG, Caloca MJ. The Rac GTPase in cancer: from old concepts to tumorigenesis. Proc Natl Acad Sci U S A 2017;114:7077–82. new paradigms. Cancer Res 2017;77:5445–51. 35. Wang J, Abate-Shen C. The MSX1 homeoprotein recruits G9a methyltrans- 16. Cao L, Yu Y, Darko I, Currier D, Mayeenuddin LH, Wan X, et al. Addiction to ferase to repressed target genes in myoblast cells. PLoS One 2012;7:e37647. elevated insulin-like growth factor I receptor and initial modulation of the 36. Ling BM, Bharathy N, Chung TK, Kok WK, Li S, Tan YH, et al. Lysine AKT pathway define the responsiveness of rhabdomyosarcoma to the methyltransferase G9a methylates the transcription factor MyoD and targeting antibody. Cancer Res 2008;68:8039–48. regulates skeletal muscle differentiation. Proc Natl Acad Sci U S A 2012; 17. Wan X, Helman LJ. Levels of PTEN protein modulate Akt phosphorylation 109:841–6. on serine 473, but not on threonine 308, in IGF-II-overexpressing rhab- 37. Battisti V, Pontis J, Boyarchuk E, Fritsch L, Robin P, Ait-Si-Ali S, et al. domyosarcomas cells. Oncogene 2003;22:8205–11. Unexpected distinct roles of the related histone H3 lysine 9 methyltrans- 18. Shern JF, Yohe ME, Khan J. Pediatric rhabdomyosarcoma. Crit Rev Oncog ferases G9a and G9a-like protein in myoblasts. J Mol Biol 2016;428: 2015;20:227–43. 2329–43. 19. Gryder BE, Yohe ME, Chou HC, Zhang X, Marques J, Wachtel M, et al. PAX3- 38. Rao VK, Ow JR, Shankar SR, Bharathy N, Manikandan J, Wang Y, et al. G9a FOXO1 establishes myogenic super enhancers and confers BET bromo- promotes proliferation and inhibits cell cycle exit during myogenic differ- domain vulnerability. Cancer Discov 2017;7:884–99. entiation. Nucleic Acids Res 2016;44:8129–43. 20. Bohm€ M, Wachtel M, Marques JG, Streiff N, Laubscher D, Nanni P, et al. 39. Clermont PL, Crea F, Chiang YT, Lin D, Zhang A, Wang JZ, et al. Identi- Helicase CHD4 is an epigenetic coregulator of PAX3-FOXO1 in alveolar fication of the epigenetic reader CBX2 as a potential drug target in advanced rhabdomyosarcoma. J Clin Invest 2016;126:4237–49. prostate cancer. Clin Epigenetics 2016;8:16. 21. Bharathy N, Suriyamurthy S, Rao VK, Ow JR, Lim HJ, Chakraborty P, et al. 40. Wang W, Chen Y, Wang S, Hu N, Cao Z, Wang W, et al. PIASxa ligase P/CAF mediates PAX3-FOXO1-dependent oncogenesis in alveolar rhab- enhances SUMO1 modification of PTEN protein as a SUMO E3 ligase. domyosarcoma. J Pathol 2016;240:269–81. J Biol Chem 2014;28:3217–30. 22. Shinkai Y, Tachibana M. H3K9 methyltransferase G9a and the related 41. Cheng HW, Chen YF, Wong JM, Weng CW, Chen HY, Yu SL, et al. Cancer molecule GLP. Genes Dev 2011;25:781–8. cells increase endothelial cell tube formation and survival by activating the 23. Shankar SR, Bahirvani AG, Rao VK, Bharathy N, Ow JR, Taneja R. G9a, a PI3K/Akt signaling pathway. J Exp Clin Cancer Res 2017;36:27. multipotent regulator of gene expression. Epigenetics 2013;8:16–22. 42. Purcell DJ, Khalid O, Ou CY, Little GH, Frenkel B, Baniwal SK, et al. 24. Chen MW, Hua KT, Kao HJ, Chi CC, Wei LH, Johansson G, et al. H3K9 Recruitment of coregulator G9a by Runx2 for selective enhancement or histone methyltransferase G9a promotes lung cancer invasion and metas- suppression of transcription. J Cell Biochem 2012;113:2406–14. tasis by silencing the cell adhesion molecule Ep-CAM. Cancer Res 2010;70: 43. Wang L, Dong X, Ren Y, Luo J, Liu P, Su D, et al. Targeting EHMT2 reverses 7830–40. EGFR-TKI resistance in NSCLC by epigenetically regulating the PTEN/AKT 25. Dong C, Wu Y, Yao J, Wang Y, Yu Y, Rychahou PG, et al. G9a interacts with signaling pathway. Cell Death Dis 2018;9:129. Snail and is critical for Snail-mediated E-cadherin repression in human 44. Tamura M, Gu J, Matsumoto K, Aota S, Parsons R, Yamada KM. Inhibition breast cancer. J Clin Invest 2012;122:1469–86. of cell migration, spreading, and focal adhesions by tumor suppressor 26. Lee SH, Kim J, Kim WH, Lee YM. Hypoxic silencing of tumor suppressor PTEN. Science 1998;280:1614–7. RUNX3 by histone modification in gastric cancer cells. Oncogene 2009;28: 45. Liliental J, Moon SY, Lesche R, Mamillapalli R, Li D, Zheng Y, et al. Genetic 184–94. deletion of the Pten tumor suppressor gene promotes cell motility by 27. Lehnertz B, Pabst C, Su L, Miller M, Liu F, Yi L, et al. The methyltransferase activation of Rac1 and Cdc42 GTPases. Curr Biol 2000;10:401–4. G9a regulates HoxA9-dependent transcription in AML. Genes Dev 2014; 46. Haddadi N, Lin Y, Travis G, Simpson AM, Nassif NT, McGowan EM. PTEN/ 28:317–27. PTENP1: `regulating the regulator of RTK-dependent PI3K/Akt signaling', 28. Son HJ, Kim JY, Hahn Y, Seo SB. Negative regulation of JAK2 by H3K9 new targets for cancer therapy. Mol Cancer 2018;17:37. methyltransferase G9a in leukemia. Mol Cell Biol 2012;32:3681–94. 47. Heller H, Gredinger E, Bengal E. Rac1 inhibits myogenic differentiation by 29. Ding J, Li T, Wang X, Zhao E, Choi JH, Yang L, et al. The histone H3 preventing the complete withdrawal of myoblasts from the cell cycle. J Biol methyltransferase G9A epigenetically activates the serine-glycine synthesis Chem 2001;276:37307–16. pathway to sustain cancer cell survival and proliferation. Cell Metab 2013; 48. Meriane M, Charrasse S, Comunale F, Mery A, Fort P, Roux P, et al. 18:896–907. Participation of small GTPases Rac1 and Cdc42Hs in myoblast transfor- 30. Kang J, Shin SH, Yoon H, Huh J, Shin HW, Chun YS, et al. FIH is an oxygen mation. Oncogene 2002;21:2901–7. sensor in ovarian cancer for G9a/GLP-driven epigenetic regulation of 49. Hennessy BT, Smith DL, Ram PT, Lu Y, Mills GB. Exploiting the PI3K/AKT metastasis-related genes. Cancer Res 2018;78:1184–99. pathway for cancer drug discovery. Nat Rev Drug Discov 2005;4:988–1004. 31. Bao L, Chen Y, Lai HT, Wu SY, Wang JE, Hatanpaa KJ, et al. Methylation 50. Janku F, Yap TA, Meric-Bernstam F. Targeting the PI3K pathway in cancer: of hypoxia-inducible factor (HIF)-1a byG9a/GLPinhibitsHIF-1 are we making headway? Nat Rev Clin Oncol 2018;15:273–91. transcriptional activity and cell migration. Nucleic Acids Res 2018; 51. Marei H, Malliri A. Rac1 in human diseases: the therapeutic potential of 6576–91. targeting Rac1 signaling regulatory mechanisms. Small GTPases 2017;8: 32. Wei L, Chiu DK, Tsang FH, Law CT, Cheng CL, Au SL, et al. Histone 139–63. methyltransferase G9a promotes liver cancer development by epigenetic 52. Helin K, Dhanak D. Chromatin proteins and modifications as drug targets. silencing of tumor suppressor gene RARRES3. J Hepatol 2017;67:758–69. Nature 2013;502:480–8.

www.aacrjournals.org Cancer Res; 79(9) May 1, 2019 2243

Epigenetic Regulation of the PTEN−AKT−RAC1 Axis by G9a Is Critical for Tumor Growth in Alveolar Rhabdomyosarcoma

Akshay V. Bhat, Monica Palanichamy Kala, Vinay Kumar Rao, et al.

Cancer Res 2019;79:2232-2243. Published OnlineFirst March 4, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-18-2676

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2019/03/02/0008-5472.CAN-18-2676.DC1

Cited articles This article cites 51 articles, 19 of which you can access for free at: http://cancerres.aacrjournals.org/content/79/9/2232.full#ref-list-1

Citing articles This article has been cited by 3 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/79/9/2232.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/79/9/2232. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.